151 3.8 Combined testing of different approaches In the previous parts we reviewed several polit-economic approaches separately. However, some of them can be combined (as especially the partisan politics approach, the opportunistic approach269 and the approach of divided government) and tested simultaneously. What do we find if we test these approaches statistically based on our data-set? We have tested variations of the following equation using OLS: RILE EL tt4t3t20 ????? ++++=? tT With: 0? , x? Constant, coefficients ?Tt Fiscal effects of tax reforms (net fiscal effects, fiscal effects of increases, fiscal effects of reductions) ELt Election year variable (dummy) LEt Left-wing government (dummy) RIt Right-wing government (dummy) EL? Stochastic error term To make tax reforms comparable over different years we have divided the fiscal effects by nominal GDP. All employed polit-economic variables are dummies. Table 10 reports our results for the dependent variables “net fiscal effects of tax reforms”, “fiscal effects of tax increases” and “fiscal effects of tax reductions”.270 We do not find strong correlations. Most of the coefficients are not significant and we never reach adjusted R2 scores higher than 0.27.271 In our results we find no indication for a partisan hypothesis. On the other hand our analysis shows that the opportunistic approach, which includes the variable for undivided government and the election dummy, performs best with respect to the fiscal effects of tax burden increases. The coefficients for both variables are highly significant at the 1% level and the adjusted R2 score is with 0.27 high compared to other specifications. As expected by the theory of opportunistic government behavior, elections had a negative effect on the extent of tax increases before elections. Governments postponed tax increases after the elections to improve their re-election probabilities. Furthermore, our observation that undivided governments refrain from 269 Frey and Schneider (1978) have early pointed out that partisan and opportunistic approaches can be combined. Parties have an ideology to which they adhere in general. But before elections they might nonetheless behave opportunistically. 270 The Durbin-Watson test statistics for our regression (reported in table 3) do not support the hypothesis of serial autocorrelation in the error terms. 271 Partly this is affected by the fact that we are focusing here on data on fiscal effects of adopted reforms as we would expect effects of partisan constellation and divided government on adopted reforms. However, with respect to opportunistic behavior data on implemented reforms would be more suitable. See the discussion of opportunistic behavior in part V.3.5. 152 tax increases is shown in the statistical analysis as well: the coefficient for the influence of undivided government on tax increases has a negative sign. The coefficients for divided government as well as for the electoral dummy equal 0.004. This indicates that elections and divided government push down the fiscal effects of tax increases by 0.4% of GDP. This finding is in line with our descriptive analysis of the data (see Figure 59 and Figure 67). Notes: t-statistics of the estimated parameters in parantheses. * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level Equation Variables R2 (adjusted) Constant Majority 2nd Chamber (Dummy) Election (Dummy) Left (Dummy) Right (Dummy) DW Observations Polit-economic approaches and tax reforms in Germany 1965-2003 Partisan/Opportunistic Net fiscal effects/ GDP 0.007 (1.15) -0.005 (-1.51) 0.009 (-1.37) -0.006 (-0.993) -0.002 2.37 39 Fiscal effects of reductions/ GDP -0.003 (-0.55) 0.001 (0.53) -0.004 (-0.671) -0.003 (0.006) -0.04 2.33 39 Fiscal effects of increases/ GDP 0.01 *** (2.71) -0.006*** (-3.29) 0.005 (-1.26) -0.003 (-0.78) 0.17 2.27 39 Net fiscal effects/ GDP 0.008 (1.16) -0.005 (-1.52) 0.0009 (0.30) -0.01 (-1.38) -0.006 (-1.02) -0.03 2.42 39 Fiscal effects of reductions/ GDP -0.002 (-0.32) -0.0002 (0.072) 0.005 (1.78) -0.007 (-1.11) -0.005 (-0.95) 0.02 2.52 39 Fiscal effects of increases/ GDP 0.25 0.01*** (2.54) -0.0054* (-2.79) -0.004** (-2.16) -0.003 (-0.76) -0.001 (-0.36) 2.05 39 Net fiscal effects/ GDP -0.00 (-0.37) -0.002 (-0.85) -0.03 2.17 39 --0.0001 (-0.054) Fiscal effects of reductions/ GDP -0.008*** (-5.04) 0.002 (0.79) 0.04 2.38 39 0.004* (1.57) Fiscal effects of increases/ GDP 0.007*** (7.45) -0.004*** (-2.71) 0.27 1.90 39 -0.004*** (-2.70) Net fiscal effects/ GDP 39 -0.00 (-0.43) -0.002 (-0.88) -0.00 2.17 Fiscal effects of reductions/ GDP 39 -0.007*** (-5.04) 0.002 (1.03) 0.001 2.28 Fiscal effects of increases/ GDP 39 0.006*** (6.60) -0.005*** (-2.96) 0.17 2.10 PartisanOpportunisticDivided Government Explanatory Variables: Own calculations based on: Federal Ministry of Finance (2004)/Federal Statistical Office(2007). Table 10: Testing positive hypotheses of tax reforms in Germany 1965-2003 Taken together our combined analysis supports our previous findings (see part V.3.3 to V.3.7) of an important effect of opportunistic behavior and divided government on the pattern of tax reforms especially via tax burden increases. 153 3.9 Summary The most important advantage of our data-set of fiscal effects of tax reforms from 1964 to 2004 is the unique opportunity to test important polit-economic hypotheses on tax policy and tax reform directly and quantitatively. Our empirical analyses showed that inertia did not play a role in tax policy in Germany. Although the revenue structure was very stable, tax reforms were very frequent and cumulated absolute fiscal effects of tax reforms over GDP between 1965 and 2003 accounted for twice the average annual tax over GDP ratio. For the second polit-economic approach – fiscal illusion by the timing of tax burden increases and reductions – we were not able to find any support either. Instead, we observed that federal governments behaved opportunistically: they tried to increase their re-election probabilities by scheduling tax burden increases directly after, and tax burden reductions in the year before the elections. Thereby governments focused on the timing of implemented reforms and not of adopted reforms. As predicted by the theory of opportunistic government behavior, wage and income taxes were most important for electoral manipulation while the timing of reforms in most indirect and less visible taxes (especially VAT and the tobacco tax) showed no sign of opportunistic behavior. While opportunistic motivation came out very clearly in our analyses, we were not able to find any empirical support for our hypotheses on differences in tax policy depending on the partisan orientation of governments. Most surprising were our findings with respect to the importance of divided government. Instead of legislative gridlock, which plays a prominent role especially in the case-study based literature on tax reform in Germany, we found by far more and higher fiscal effects of tax reforms under divided than under undivided government. In our data there was no indication that a majority of the opposition effectively blocked tax reforms, nor that tax reforms were larger and more frequent in case of undivided government. Furthermore, there was no evidence that divided federal governments were concentrating on reforms of taxes which did not need approval of the second chamber. Finally, there was only very limited evidence for a stronger role of partisan preferences in case of undivided government. Instead, we found that especially tax burden increases were substantially lower in case of undivided governments. This shows that especially undivided governments restrained from tax burden increases as they feared to be blamed exclusively for tax burden increases in the next elections.